BillBxa is an AI-powered platform that streamlines utility and telecom bill payments for enterprises, enabling bulk processing, real-time tracking, and reconciliation via a secure dashboard.
BillBxa is an AI-powered platform that ingests utility and telecom bills in bulk, extracts and categorises their data, pays them through connected gateways and reconciles the results for enterprise finance teams. The backend is a set of Java Spring microservices on AWS, with each stage of the bill lifecycle (ingest, parse, pay, reconcile) running as an independently scalable service joined by queues. That shape fit a product whose load arrives in spiky batches of thousands of bills and whose parsing step is slow and AI-bound while payment and tracking must stay responsive.
A Flutter mobile app for scanning and uploading bills, plus an enterprise web dashboard for bulk operations and real-time status.
A single authenticated entry point that terminates TLS, enforces tenant scoping and rate limits, and routes requests to services.
Java Spring microservices, one per stage of the bill lifecycle, communicating through asynchronous jobs so a slow parse never blocks a payment.
Relational storage for bills, payments and ledgers; object storage for original bill documents; queues to decouple stages.
An OpenAI-backed extraction pipeline turns 40+ bill formats into structured fields, and adapters connect to utility providers and payment gateways.
Containerised services on AWS with autoscaling workers for parsing batches, centralised logging and encrypted secrets.
An enterprise user uploads a batch of bills (PDF, image or email forward) from the Flutter app or web dashboard.
The gateway authenticates the tenant and the ingestion service stores originals in object storage, emitting one parse job per bill.
Parsing workers call the OpenAI extraction pipeline, validate amounts, due dates and account numbers, and write structured records.
Approved bills are queued to the payment service, which calls the right utility provider or payment gateway and records the response.
Gateway webhooks and provider callbacks update payment status; the tracking service pushes live updates to the dashboard.
The reconciliation engine matches payments against statements, flags anomalies, and finance teams export reports or get WhatsApp alerts.
Parsing is CPU and API-latency heavy while payment and tracking need low latency; splitting them lets parsing workers scale independently during batch uploads without affecting dashboards.
Thousands of bills arrive at once; placing a queue between ingest, parse, pay and reconcile absorbs spikes, gives retries for flaky provider APIs, and makes every step idempotent and auditable.
An OpenAI-based extractor handled hundreds of bill layouts without hand-written templates, but every extracted field passes schema and business-rule validation before any money moves.
Utility providers and payment gateways each have different APIs and failure modes; a uniform adapter interface let the team add new providers without touching the core payment flow.
We ship MVPs, AI agents, and production systems in 8 weeks. Bring the rough spec — we'll scope the build inside a week with a fixed timeline and quote.
